Foreign Aid, Foreign Policy, and Domestic
Government Legitimacy: Experimental Evidence
from Bangladesh
Simone Dietrich, Minhaj Mahmud, Matthew S. Winters
Abstract:
Foreign aid donors try to make themselves visible as the funders of development projects in order to improve
citizen attitudes abroad. Do target populations receive these political communications in the intended fashion,
and do they succeed in changing attitudes? Despite the widespread use of the practice, there exists little evidence
about the effectiveness of this strategy. We embed an informational experiment about a U.S.-funded health
project in a nationwide survey in Bangladesh. Although we find limited recognition of the USAID brand, explicit
information about U.S. funding slightly improves general perceptions of the United States. It does not, however,
change respondent’s opinions on substantive foreign policy issues. We also find, contrary to existing arguments
that foreign aid undermines domestic government legitimacy, that the information increases confidence in local
authorities. These results strengthen our understanding of the efficacy of promoting donor visibility and shed
light on an important debate in the area of governance that assesses the effect of external actors on government
legitimacy.
Keywords: Foreign aid, aid branding, information experiment, Bangladesh
Working Paper 16
December 2015
Foreign Aid, Foreign Policy, and Domestic Government
Legitimacy: Experimental Evidence from Bangladesh
Simone Dietrich
University of Essex
Correspondence to:
[email protected]
Simone Dietrich is Senior Lecturer of Government at the University of Essex. She is also Director of ESSEXLab,
the University’s social science experimental lab. Prior to joining Essex, Simone was Assistant Professor
of Political Science at the University of Missouri, Columbia. Her research is published or forthcoming in
International Organization, Journal of Politics, International Studies Quarterly, Journal of Experimental
Political Science, World Development, European Journal of Development Research, and Oxford University
Press.
Minhaj Mahmud
BRAC Institute for Governance and Development
BRAC University
Correspondence to:
[email protected]
Minhaj Mahmud is currently head of research of BRAC Institute of Governance and Development, BRAC
University. Previously he has held various academic and research positions at Queens University Belfast and
Keele University in UK, Bangladesh Institute of Development Studies (BIDS) and Jahangirnagar University
in Bangladesh. He has completed his Ph.D. in Economics from the University of Gothenburg. His research
broadly focuses in behavioral economics, experimental economics and development issues. His research
is published in journals such as Economics Letters, Journal of Economic Behavior and Organization,
Economica and Health Economics. He is currently a Visiting Scholar at the JICA Research Institute, Tokyo.
Matthew S. Winters
University of Illinois
Correspondence to:
[email protected]
Matthew S. Winters is Associate Professor in the Department of Political Science at the University of
Illinois at Urbana-Champaign. He received his Ph.D. in Political Science from Columbia University and
was a Postdoctoral Research Fellow at the Niehaus Center for Globalization and Governance at Princeton
University. His research interests include the political economy of foreign aid, corruption, and the processing
of political information. He has published articles in Comparative Politics, International Studies Quarterly,
World Development, and World Politics, among other outlets.
Working Paper 16
December 2015
Acknowledgments:
AidData:
[email protected]
www.aiddata.org
Working Paper 16
December 2015
Versions of this paper have previously been presented at the 2014
International Political Economy Society Annual Meeting; the Brown
University Conference on Accountability in Low and Middle Income
Democracies; the SFB 700 and AidData Workshop on Foreign Aid and
Effective and Legitimate Governance in Areas of Limited Statehood;
the 2015 International Studies Association, European Political Science
Association, and American Political Science Association annual meetings;
the ISA and AidData Workshop on Foreign Aid and Government Legitimacy
at Georgetown University; the 2015 Center for Advanced Studies Spring
Symposium at the University of Illinois; the Niehaus Center for Globalization
and Governance Conference on The Politics of Government Revenue at
Princeton University; Northwestern University; the University of Pittsburgh;
and Yale University. Thanks to David Bearce; Rob Blair, Ceyda Erten,
Raymond Hicks, Susan Hyde, Nolan McCarty; Helen Milner, Laura Paler,
Steven Radelet, and Kelly Senters for excellent discussions or comments
on previous versions. Thanks to Saumil Dharia for excellent research
assistance. We are grateful for financial support from the University of
Missouri Research Board and the International Studies Association.
AidData – a joint venture of the College of William and Mary, Development
Gateway and Brigham Young University – is a research and innovation lab
that seeks to make development finance more transparent, accountable,
and effective. Users can track over $40 trillion in funding for development
including remittances, foreign direct investment, aid, and most recently
US private foundation flows all on a publicly accessible data portal on
AidData.org. AidData’s work is made possible through funding from and
partnerships with USAID, the World Bank, the Asian Development Bank,
the African Development Bank, the Islamic Development Bank, the Open
Aid Partnership, DFATD, the Hewlett Foundation, the Gates Foundation,
Humanity United, and 20+ finance and planning ministries in Asia, Africa,
and Latin America.
Contents
Abstract ........................................................................................................................................................ 1
1. Introduction .............................................................................................................................................. 5
2. How Does Information about Foreign Aid Change Citizen Attitudes? ...................................................... 8
3. Potential Externalities of Information about Foreign Aid on Domestic Government Legitimacy ............ 10
4. Research Design.................................................................................................................................... 11
4.1 Experimental Interventions .............................................................................................................. 11
5. Outcomes ............................................................................................................................................... 13
6. Results .................................................................................................................................................. 14
6.1 H1: Brands and Logos Spread Information about Donor Funding .................................................. 15
6.2 H2: Effects of Information about Foreign Aid on Attitudes toward the Donor .................................. 17
6.3 H3: Effects of Information about Foreign Aid on Domestic Government Legitimacy ...................... 20
7. Conclusions ........................................................................................................................................... 23
Tables and Figures .................................................................................................................................... 26
Appendix .................................................................................................................................................... 33
References ................................................................................................................................................. 43
4
1. Introduction
“[T]he branding … absolutely had a major impact on their perception of the U.S. I think our new global
branding is a major foreign policy achievement.” – Bill Frei, USAID (ND: 3)
In the introduction to their book on anti-Americanism, Katzenstein and Keohane (2007) outline three
reasons why improving attitudes toward the United States around the world is fundamental to U.S. foreign
policy goals. First, in the age of global terrorism, negative opinions toward the United States help to
create “breeding grounds for terrorism” in which terrorist organizations can recruit activists and draw on
the resources of sympathizers.
Since there may be direct security consequences of high levels of
negative affect toward the United States, U.S. policymakers have an interest in cultivating positive
sentiment among foreign populations. Second, the general character of public opinion is relevant for
international cooperation, an ever more pressing objective in the era of climate change, the globalization
of disease, and other transnational threats.
Positive public attitudes toward potential international
partners may be an important facilitator of interstate cooperation. Third and relatedly, the least expensive
power to exercise is “soft power,” the power that relies on others wanting what you want and trusting in
your judgment (Nye 2002). Where positive affect toward the United States exists, the United States will
be able to achieve diplomatic and security outcomes at lower cost.
When giving foreign aid, donor governments conduct public diplomacy by directly communicating their
financing role to intended beneficiaries abroad. An example can be found in Bagerhat District,
Bangladesh, where there is a large sign on the side of a newly built school building. The sign depicts two
women building a cyclone shelter along the side of a mangrove tree. Clockwise from the upper-right-hand
corner of the sign, an observer can find logos for three different nonprofit development organizations
headquartered in the United States or the United Kingdom (ACDI/VOCA, PCI, and MuslimAid) and for the
U.S. Agency for International Development (USAID), the overall funder of the project.
This type of
``branding” is the most common way by which donors spread information to populations in developing
countries about their role in funding public goods and social services.
This type of public diplomacy has a long history in the United States as being a means for promoting the
country’s soft power by influencing public opinion. Nye (2008: 97, citing Pellis 1997: 33) notes that the
Roosevelt administration viewed America’s security to depend “on its ability to speak to and win the
support of people in other countries.” Such public diplomacy has played a central role in foreign aid policy
since the Foreign Assistance Act of 1961, which has required that U.S. overseas projects be marked as
“American Aid” in order to win the loyalties of local populations in the developing world. This effort at
5
spreading information about U.S. development efforts was redoubled as part of the post-9/11 National
Security Strategy. According to Andrew Wilder (2010), U.S. foreign aid in Afghanistan, Pakistan, and Iraq
has been used as a ‘weapons system’ to promote stabilization and counterinsurgency objectives” (407).
1
By communicating the origin of sponsorship to local populations, the U.S. government seeks to establish
positive associations with and affect toward the United States among people in need.
Other donor
countries, such as the United Kingdom, have followed suit with their own aggressive branding policies.
Despite these prominent efforts by powerful countries to change attitudes around the world, little scholarly
attention has been paid to the effectiveness of public diplomacy in changing opinions and influencing
2
policy abroad from the grassroots level. What is more, little attention has been paid to the potential
unintended consequences that may arise from public diplomacy efforts and that may ultimately undermine
the strategic objectives of the United States or other donor countries. For instance, there is a prominent
theoretical argument that information about the external funding of development interventions will
undermine citizen support for their own governments where citizens expect the state to provide goods and
services (Bratton 1989; Fowler 1991; Whaites 1998; Gubser 2002; Brass 2010). Such a disconnect
between citizens and the state could lead to lower levels of government stability and reduce the quality of
governance in a state in the long run (Brautigam 2000; Knack 2001; Hoffman and Gibson 2005).
3
To date, little rigorous evidence exists about the consequences of spreading information about foreign
development financing on peoples’ attitudes towards the donor. Nor do we know whether the most
commonly used strategy for spreading such information – the branding of development interventions
through the use of an aid agency logo – even effectively communicates that a foreign donor has
4
sponsored a given intervention.
In their publicity materials, USAID cites macro-level observational data
on changed attitudes toward the United States in Indonesia following provision of humanitarian relief in
the wake of the December 2004 Indian Ocean tsunami.
5
But the persistence of strong anti-American
1
According to Wilder (2010), foreign aid has been codified in U.S. Army counterinsurgency policy, which now includes non-military
public affairs operations like “publicizing the opening of a water treatment facility” because such public goods, when labeled as being
from the U.S. government, are thought to facilitate the overarching objective of winning the support of citizens away from insurgents
and in favor of the foreign military and the government. Berman, Shapiro, and Felter (2011) use panel data to show that
reconstruction spending in Iraq reduced insurgent violence. See also Lyall and Wilson (2009).
2
Existing scholarship largely focuses on government-to-government influence, where foreign aid is leveraged to convince aidreceiving governments to make policy concessions or promote democratic change at home. Recent examples in this tradition
include Steinwand (2015, 2014), Dietrich and Wright (2015), Vreeland and Dreher (2014), Swedlund (2014), Bearce (2013), Bearce
and Tirone (2010), and Flores and Nooruddin (2009).
3
Other scholars, however, have downplayed this alleged risk to state legitimacy and stability posed by foreign aid. Successful
management of foreign aid relationships may actually legitimate the aid-receiving state when citizens observe representatives of the
government working with non-state actors in a collaborative fashion (Tendler 1997).
4
Beyond citizens in developing countries, branding may also target donor publics and donor legislatures, serving as an important
tool to communicate to domestic audiences what foreign aid efforts accomplish. Branding increases transparency by providing visual
evidence of specific development contributions (Shah 2010)
5
Bill Frei, the Director of USAID’s Indonesia program at the time, argued, “The people of [Aceh] (Indonesia) saw the branding; they
knew right away the U.S. government was responding. That absolutely had a major impact on their perception of the U.S. Without
the branding, USAID would have been just one of 550 NGOs working in the area” (USAID ND: 3). See also
http://www.usaid.gov/branding (accessed 26 May 2015).
6
sentiment in countries to which the United States sends significant amounts of assistance raises
questions about whether information about U.S. development activity is having the intended effect, as do
not-infrequent official requests for USAID to avoid branding its interventions (Adelman 2011; MerkovicOrenstein 2011; Reid-Henry 2011; Birdsall, Vaishnav, and Cutherell 2012).
6
Using an original informational experiment in Bangladesh that includes a number of innovative outcome
measures, we shed light on three important questions: Do conventional foreign aid branding techniques
effectively communicate that a project was funded by a foreign country? Does information about donor
sponsorship improve attitudes toward the donor? Does this information undermine the legitimacy of the
aid-receiving government in the eyes of its own citizens?
We also explore whether the effects of
information about foreign aid vary across different types of people (e.g., those who have had greater
previous exposure to the project; those who are more politically sophisticated).
We research these
questions with reference to a prominent network of health clinics in Bangladesh – the Smiling Sun Clinics
– that are partially funded by USAID.
A positive example of what foreign aid can do, these clinics
7
generally are regarded as quite successful in delivering healthcare to the population that they serve.
Bangladesh serves as a useful testing ground for studying the effects of information about foreign aid. At
the level of government-to-government relations, Bangladesh has long served the United States as an
important partner in its global fight against terror (Congressional Research Service, 2010). The country is
the third largest Asian recipient of U.S. assistance after Afghanistan and Pakistan (State Department
8
2015) , and U.S. government officials have emphasized the dual role that foreign aid can play in
improving the country’s counterterrorism capacity while also promoting good relations between the two
countries (Biswal 2014).
These positive government-to-government relations, however, mask growing anti-Westernism among the
population. In 2015, Italian and Japanese aid workers were killed in the country, and a string of attacks
9
against secular bloggers and publishers left more than a half-dozen Bangladeshis dead (Anam 2015).
Despite the long history of good relations between Bangladesh and the United States and while the
country is not known for anti-American demonstrations of the type found in Iran or Pakistan, there
nonetheless appears to be a role for public diplomacy to play in Bangladesh as a relevant tool for
reducing anti-Western sentiment.
6
Recent papers by Nooruddin (2014) and Hyde (2015) highlight the potential of survey experiments for studying (sensitive)
questions in international relations and comparative politics.
7
See Lance, Angeles, and Kamal (2012) for an impact evaluation of the clinics.
8
From all bilateral and multilateral donors, Bangladesh received US$17 per capita in foreign development assistance in 2014; this
amount has been steadily increasing over the last 10 years (World Development Indicators 2014).
9
For more on attacks against Western aid workers, see Narang (2013).
7
We find that the conventional U.S. practice of marking aid projects with the USAID logo is relatively
ineffective at revealing information about U.S. sponsorship. Once the role of the United States as financier
is made explicit, however, Bangladeshis who receive the information are more likely to view the U.S. in a
positive light as compared to other citizens who do not receive the information about U.S. sponsorship.
We find that treatment effects are heterogeneous and vary by population sub-groups. We find less
evidence that the information changes any specific policy attitudes. These findings suggest both that
public diplomacy through foreign aid is possible but also that more thought may need to be put into how to
spread relevant information.
In what many will regard as a surprising finding, we find that – rather than delegitimizing the domestic
government – information about donor financing improves our respondents’ views of their local
governments. In other words, citizens reward their own governments for bringing in external funding for
development interventions. Again, we identify important heterogeneous treatment effects across subgroups. Our findings directly contribute to an emerging debate among comparative politics and
development scholars about the effects of foreign aid on governance, as well as to our general
understanding of how citizens perceive the quality of governance and give credit to different levels of
government.
2. How Does Information about Foreign Aid Change Citizen
Attitudes?
Despite donor governments’ assertions that information about foreign aid sponsorship can improve the
attitudes of citizens in aid-receiving states toward the donor, to the best of our knowledge, no foreign aid
agency has ever fully outlined a relevant theory of change. We understand the logic underlying the
potential benefits of information about incoming foreign aid flows to originate in the notion that
development interventions are likely to be perceived as helpful and beneficial by local populations. The
spread of information about the sources of funding therefore should lead to the transfer of positive affect
about the development intervention itself to the donor country that has funded it.
If the mechanism of information transmission is the branding of foreign aid with the donor agency logo –
as it most commonly is – we must assume that people associate the donor brand (e.g., the USAID logo)
with the donor government. This assumption seems to have been treated relatively uncritically in the
development world. Given low levels of literacy and education, it is far from certain that people who see a
donor logo will necessarily recognize that it is associated with a far-away country or – even more to the
point – that it is advertising foreign sponsorship of an aid project. That is, the target audience of foreign
8
aid branding may not actually infer that the logo has been placed on a development project to indicate
financial sponsorship or the transfer of resources. In the empirical section below, we test the initial
hypothesis that donor branding increases citizens’ awareness of foreign sponsorship of a development
intervention (H1). Our results show that this hypothesis is confirmed for only a limited proportion of
respondents.
We subsequently test the main hypothesis related to the effects of information about foreign aid. When
explicitly informed about the aid sponsor,
10
citizens may update their opinions of and attitudes about the
donor in a positive fashion through either conscious or subconscious processing of the informational
treatment. At a conscious level, we can imagine citizens reflecting on the fact that a foreign donor has
provided an intervention in a way that generates cognized feelings of gratitude that lead to more positive
affect toward the donor. At a subconscious level, it is possible that positive attitudes about or valuations
of a development intervention are subconsciously transferred to the project sponsor through the process
that Lodge and Taber (2013) refer to as “hot cognition.” Research on “affect heuristics” suggests that
current emotion influences decisions (e.g., Slovic and Peters 2006; Wilson and Arvai 2006), such that
positive emotions generated by information about foreign sponsorship of a development intervention
should then make people more likely to judge the U.S. in a more positive light and to be more supportive
of U.S. policies (H2).
11
Previous results in the literature are mixed. Using macro-level data, Goldsmith, Horiuchi, and Wood
(2014) show that variation in levels of funding from the United States’ President’s Emergency Plan for
AIDS Relief (PEPFAR) positively predict attitudes toward the United States across about 50 PEPFARreceiving countries.
Andrabi and Das (2010) show that attitudes toward foreigners vary based on
proximity to the fault line of a 2005 earthquake in northern Pakistan that brought a steady flow of
humanitarian aid to the country. A study of foreign assistance in northeast Afghanistan over the period
2007-2009 reveals that foreign aid “did not have an impact on attitudes toward foreign forces” (Zürcher,
Koehler, and Böhnke 2010, 5), while qualitative work from Afghanistan reports that “aid and development
projects were consistently negatively described by Afghans” (Fishstein and Wilder 2012, 41). In survey
data collected from Indian respondents, Dietrich and Winters (2015) find a small but not statistically
significant effect of information about U.S. sponsorship of a foreign aid project in the health sector on
respondents’ perceptions of U.S. favorability.
10
In the treatment that we employ in our experiment, we directly provide information about sponsorship after exposing respondents
to the brand and asking them an initial follow-up question. This allows us to investigate whether the brand is conveying the intended
information and then investigate the effects of the actual information that donor brands are intended to convey.
11
Other affective associations are possible. If the reaction to a development intervention is negative (e.g., because of poor project
implementation or cultural inappropriateness), this affect may be consciously or subconsciously transferred to the foreign donor.
9
3. Potential Externalities of Information about Foreign Aid
on Domestic Government Legitimacy
The distribution of information about external funding of development projects may also influence citizens’
views of their own government.
12
Some have suggested that widespread information about the presence
of donors in various social and economic sectors may signal a lack of competence or willingness on the
part of the government to provide basic goods and services, rendering governments unable to make
legitimate demands of their citizens (Bratton 1989; Fowler 1991; Whaites 1998; Gubser 2002; Brass
2010). For instance, external funding of basic goods and services may undermine the “fiscal contract”
between the state and its citizens, lowering the tax morale of citizens. This potential disconnect may
ultimately contribute to a pattern in which foreign aid flows undermine the quality of governance in a state
(Brautigam 2000; Knack 2001; Hoffman and Gibson 2005). Insofar as foreign aid is given for the purpose
of supporting allied governments, this undermining of an ally’s domestic legitimacy might overshadow any
benefits of improved perceptions of the donor brought about by the diffusion of funding information.
On the other hand, successful management of relationships with foreign aid donors may actually help
legitimize the state and its institutions (Tendler 1997).
Since many donor-funded development
interventions directly target some local communities but not others and since project implementation
usually requires the consent and cooperation of local authorities, citizens might directly credit their local
authorities for bringing a project into their community. Indeed, the presence of externally-funded projects
can serve as signals of the competence of a local government in providing goods and services to its
community (Cruz and Schneider forthcoming; Guiteras and Mobarak 2014). If citizens perceive local
leaders as playing a role in the project negotiation process it is plausible that leaders will receive credit for
securing the project for the local area. What is more, local authorities often run on election platforms
where they credit themselves with the promotion of local development.
We therefore have two competing hypotheses. It may be that information about donor sponsorship of
development interventions undermines domestic government legitimacy (H3a). Alternatively, it may be
that information about donor sponsorship of development interventions improves citizen perceptions of
their domestic government (H3b).
So far the literature has found little evidence of a negative externality and more often has estimated a
positive relationship between non-state service provision and attitudes toward the state. Using data from
12
Increasingly, scholarship is conducting individual-level analyses to understand the influence of external actors on peoples’
perceptions of their domestic institutions. Examples include Bakke (2015) and Blair (2015). Carnegie and Dolan (2015) propose that
some governments will refuse foreign aid in order to protect their public image.
10
the Afrobarometer surveys, Sacks (2012) finds a positive correlation between non-state service provision
and tax morale (i.e., citizens’ willingness to recognize the state’s legitimate right to collect taxes). This is
also the finding of Zürcher, Köhler, and Böhnke (2010) in northeastern Afghanistan: in places with higher
levels of foreign-funded small infrastructure projects, levels of support for the district and provincial
government were higher, although this effect did not persist across the four years of the study (see also
Böhnke and Zürcher 2013). Guiteras and Mobarak (2014) show in a field experiment that subsidies for
small-scale infrastructure in rural Bangladesh initially increases citizen support for the local government
until people are given information that the subsidies came from an NGO and that the local government
played no role in obtaining them. Cruz and Schneider (2014) show how local government officials in the
Philippines reap electoral benefits from the presence of a World Bank-funded and central governmentimplemented development project. In survey experimental data from India, Dietrich and Winters (2015)
find no evidence that favorability ratings for domestic government officials or institutions decrease when
individuals receive information about the foreign funding of a development project.
4. Research Design
To test our hypotheses, we ran an informational experiment, embedded in a nationally representative
household survey of Bangladeshi citizens. Between 7 September and 1 October 2014, enumerators from
the BRAC Institute of Governance and Development conducted face-to-face interviews with 2,294
respondents from 55 subdistricts (upazila) spread across all seven divisions of Bangladesh.
13
During the
survey, we provided information about a development intervention to both our treatment and control
group; the treatment group received additional information about U.S. sponsorship of the intervention.
The information that we provided in the survey closely approximates the kind of information that citizens
typically would receive about donor sponsorship (e.g., through the media or through conversations in their
communities).
4.1 Experimental Interventions
After two initial questions about what policy issues were of importance to the respondents and the current
state of affairs in Bangladesh, we assigned randomly sampled respondents to one of the two versions of
14
the survey.
Both versions of the survey described and showed a brief video related to the Smiling Sun
Clinics, a network of non-profit health clinics catering to poor communities and partially funded by USAID
13
The sampling strategy is described in the Appendix.
In Table A1 in the Appendix, we show that our treatment and control groups are balanced on a number of background covariates,
including previous use of and awareness of the Smiling Sun Clinics that we use in our intervention.
14
11
since 1997. Before showing respondents the video, we asked them if they were familiar with the Smiling
Sun Clinics and if they, a family member, or a close friend had ever been to a Smiling Sun Clinic. The
video was one-minute long and dramatized the way in which the pre-natal services provided by the clinics
calmed the anxieties of two expecting parents. It was shown to respondents on the hand-held tablets on
which enumerators conducted the survey. In the control version of the survey, the video was branded at
the bottom of the screen with the Smiling Sun logo and name. In the treatment version of the survey, the
video was branded with the USAID/Bangladesh logo and the tagline (in Bangla) “On behalf of the
American People.”
Figure 1 provides screen shots from two difference scenes in the video for the
treatment and control groups; except for the branding, the video was identical in the two conditions.
Immediately following the video, we asked respondents in both conditions an open-ended question about
where “most of the financial assistance for the Smiling Sun Clinics” has come from. We use this question
to examine the first hypothesis that donor branding signals the identity of the foreign sponsor.
To
examine the lasting impression left by the donor brand, we also asked our respondents toward the end of
the survey if they could describe the logo that they had seen in the video.
In the treatment condition only, we provided direct information that the Smiling Sun Clinics are funded by
a foreign donor, informing respondents:
Since 2007, the United States has been giving assistance to build Smiling Sun Clinics in
Bangladesh. The funds have been provided through USAID. The United States has
provided more than 4,500 million taka
15
in support of the Smiling Sun Clinics.
In this condition only, we then asked our respondents if they knew why the United States was providing
financial assistance to Bangladesh and also if it was good for Bangladesh to take money from the United
States for the Smiling Sun Clinics and why. In addition to collecting useful information, these questions
were meant to strengthen the treatment, increasing the probability that respondents both absorb the
factual information that the United States had provided the funding and also think through the relationship
between the United States and Bangladesh in a way that would activate their own sentiments about the
donor-recipient relationship.
15
This is about US$57 million.
12
5. Outcomes
The remainder of the survey included outcome measures related to Bangladeshi relations with the United
States, confidence in Bangladeshi government institutions, and the quality of the Smiling Sun Clinics. We
use a mixture of questions to get at each of these topics. One of the contributions of our survey is that, in
addition to general measures of abstract attitudes toward the foreign aid donor, we use policy-specific
measures and a set of questions that judge whether affect toward the donor might spill over into attitudes
toward its commercial goods.
To measure how the treatment affects citizens’ attitudes toward the donor, we ask a straightforward
question about the influence of foreign countries on Bangladesh. Respondents were first asked about
whether the United States had “a lot” of influence on Bangladesh or “not much” influence on Bangladesh.
We asked about the United States as one of a set of ten countries. For respondents who said that the
United States had a lot of influence, we subsequently asked whether this influence was good or bad.
In addition to this general question, we asked policy-specific questions in the realm of trade and security.
If information about foreign aid is to work as public diplomacy, any warm glow that it generates needs to
also lead citizens exposed to the information to update their policy preferences on issues relevant to the
donor. For trade, we generally described U.S.-Bangladeshi trading relations and then asked whether
respondents thought that Bangladesh should give more priority to continuing trade with the United States
or rather to diversifying its trading partners.
For security, we described Bangladesh’s frequent
participation in U.N. Peacekeeping Operations and then asked respondents which of nine countries they
would most like to see in command of Bangladeshi troops.
By asking questions about specific policies, we increase the costs of showing favoritism toward the United
States. Answering the influence question is cheap and given the nature of the treatment, the “right”
answer might be obvious.
Taking a stand on trade policy or attitudes toward U.N. Peacekeeping
Operations, however, involves expressing favoritism toward the United States in an arena where the
respondent may have different initial instincts.
Beyond these questions about Bangladesh’s foreign policy, we asked two types of questions about
international commerce, looking to see if positive attitudes generated by information about foreign aid spill
over into the consumer preferences of citizens in aid-receiving countries. Existing evidence on product
marketing shows that overall beliefs and judgments about a foreign country influence individuals’
perceptions and judgments of consumer products associated with that country (Cattin, Jolibert, and
13
Lohnes 1982; Peabody 1985; Han and Terpstra 1988). If information about foreign aid is successful at
improving people’s attitudes toward the donor, this should be visible in people’s attitudes toward things
associated with the donor, such as multinational corporations headquartered in the donor country or
goods produced in the donor country.
First, we asked respondents to express their level of trust in seven different multinational corporations,
two of which are U.S.-based corporations (and were identified as such in the question prompt): Apple and
Coca-Cola. Second, we asked respondents about six different types of consumer goods and asked them
to choose which country (from a list of nine) produced the highest quality products. If positive affect is
being generally transferred to the donor country, we should see that transfer reflected in respondents’
perceptions of American companies and products.
To measure domestic government legitimacy, we asked a series of nine questions about institutions in
Bangladesh ranging from the “local government” and the “national government” to “NGOs” and “banks
and financial institutions.” For each institution, we asked respondents to say whether they had full, partial,
or no confidence in the institutions and the people leading them. We study some of these questions
individually and also an index of responses to the questions about government institutions. To parallel
the key outcome variable in Sacks (2012), we also included a tax morale question on the survey, asking
respondents to agree or disagree with the statement that “the tax department of the government has the
right to make people pay taxes.” And finally we asked respondents about levels of corruption in the
government to see if information about U.S. involvement would lead people to perceive their government
as more corrupt because it did not itself implement the project or whether U.S. involvement would
encourage people to perceive their government as less corrupt since foreign donors were willing to
operate in the country.
6. Results
We draw our inferences about the way in which donor visibility affects people’s attitudes by looking for
differential responses across these variables between the group that was told about and asked to
consider the fact that the United States funds the Smiling Sun Clinics and the group that was not exposed
to information about U.S. funding of the clinics. Throughout the paper, we report results from differencein-means tests and Wilcoxon rank-sum tests. In Table A2 in the Appendix we present results from logistic
and ordered logistic regression models that control for a host of background covariates. These results are
almost identical in terms of magnitude and statistical significance to those presented below. We further
examine heterogeneous treatment effects across certain sub-groups in the population. For example, we
14
provide sub-group analyses related to levels of education and political involvement as we expect these
variables to influence the treatment effects. We also distinguish among respondents who were previously
familiar with the Smiling Sun Clinics (and thus more likely to have been previously exposed to USAID
branding) and those who were not familiar with the intervention.
6.1 H1: Brands and Logos Spread Information about Donor Funding
Donor governments commonly rely on visual logos to spread information about donor financing. The U.S.
government uses the USAID brand to communicate its financial assistance to beneficiaries abroad. In this
section, we provide evidence that surprisingly few respondents in the treatment condition were able to
correctly infer the meaning of the USAID brand. We find that that the treatment effects are larger among
respondents with previous exposure to the clinics or higher levels of education.
In both the treatment and control conditions, respondents were asked, immediately after watching the
video and before any additional information, “Do you know from where the Smiling Sun Clinics get most of
their financial assistance?” Despite the fact that the treatment condition used a video where the USAID
logo and the statement “On behalf of the American people” (in Bangla) were present across the bottom of
the video for the duration of the video, the modal answer in both the treatment and control conditions, as
can be seen in Table 1, was for the respondent to say that he or she did not know the source of funding
for the Smiling Sun Clinics.
Although the effect size is smaller than we expected a priori, the USAID branding in the treatment version
of the video alerted at least some of the respondents to the fact that the United States was funding the
intervention. The proportion of respondents saying that the Smiling Sun Clinics were funded by the United
States increased by 13 percentage points between the control and treatment conditions. Most of the
movement came from people who otherwise would have said that they did not know. An additional threeto-four percent of respondents in both conditions said that the funding came from “foreign aid” or “foreign
countries.”
The remainder of Table 1 shows the patterns in the data for several select subgroups. We find larger
effects among people who have previously used or were previously aware of the clinics. These are
respondents who we can think of as having had multiple exposures to the treatment. Indeed, looking only
at respondents in the control group, among those who say that they had previously used the Smiling Sun
Clinics, 26 percent were able to correctly identify the United States as the funder, which is statistically
significantly different from the 8 percent of non-previous users in the control group who are able to do so
15
16
(p < 0.01).
Among previous users, the stimulus that we provided in the treatment condition increases
the proportion that cites the United States as the sponsor of the clinics by 14 percentage points (p < 0.01).
Among those who were previously aware of the clinics (but had not personally used them or had family or
friends use them), only 16 percent of the control condition respondents were correctly able to identify the
United States as the funder, a proportion that increased by 18 percentage points in the treatment
condition (p < 0.01). These effects suggest that donor brands will be most effective at informing or
reminding people of donor financing when those people have a pre-existing awareness of or connection
to that intervention.
We also looked at highly educated respondents (i.e., respondents with a completed secondary education
or above). We find the largest effect of the branding treatment among this group: a 23 percentage point
increase over a relatively high baseline of 21 percent naming the United States in the control condition.
As one might expect, more highly educated respondents seem the most capable of fully absorbing the
meaning of a brand – perhaps simply because their literacy allowed them to read the “On behalf of the
American people” tagline – and drawing a correct inference about its relationship to financial sponsorship.
Nonetheless, 47 percent of our highly educated respondents in the treatment group still said that they did
not know from where the Smiling Sun Clinics had received their funding.
Toward the end of the survey, we reminded respondents that there had been a small logo in the video
that they had watched earlier, and asked them if they could tell us what that logo had indicated. In the
treatment condition, five percent of respondents mentioned USAID in their answer, while less than one
percent of respondents in the control condition did.
In both conditions, the vast majority of people
responded to the question by talking about the Smiling Sun Clinics in general. Often, they described the
video in general and not the logo per se. In both conditions, 27 percent of respondents said that they did
not remember what had been in the video or did not know how to describe the logo that they had seen.
This is an additional piece of evidence that, insofar as the USAID logo was observed in the treatment, it
did not register with our respondents in a profound and lasting way.
These results suggest that, overall, the USAID brand is not well known in Bangladesh or is not
understood as a signal of U.S. financing. This impression was confirmed by author field visits to Smiling
Sun Clinic locations where people nearby were unable to identify the funder or to say what the USAID
logo on the clinic meant. After the tagline was read to these individuals, they were able to say that money
for the clinics must have been coming from the American people.
16
The proportion of previous clinic users in the control group who correctly identify the United States as the funder of the clinics is, in
fact, larger than the proportion of non-previous users in the treatment group who do so (δ = 0.06; p < 0.13). In Appendix Table A2,
we provide a logistic regression analysis of correct answers to the funding question among respondents in the control group.
16
6.2 H2: Effects of Information about Foreign Aid on Attitudes toward
the Donor
To test if information about U.S. sponsorship of the Smiling Sun Clinics changes Bangladeshi opinions
about the United States, we now proceed to evaluate the combined effects of the branding and the
second-stage of our treatment where we directly informed respondents in the treatment condition that the
U.S. government has been funding the Smiling Sun Clinics for almost 20 years (after having asked the
questions above where we tried to assess whether they understood the meaning of the USAID logo).
Although we did not ask respondents an additional question to confirm that they understood the explicit
information about the history of U.S. funding for the Smiling Sun Clinics, we asked respondents later in
the survey to identify which of seven foreign donor countries or international financial institutions it would
be best for Bangladesh to ask to financially support the Smiling Sun Clinics in the future. Among those
who had a preference, 65 percent in the treatment condition said the United States, while only 52 percent
named the United States in the control condition, a difference that is statistically significant (p < 0.01).
17
This suggests that a subset of respondents in the treatment condition were absorbing the information,
18
although the magnitude of the change is still smaller than we would have expected ex ante.
We also
tried to reinforce the information by asking respondents why they thought the U.S. provided such funding
and whether or not they thought that it was good that the U.S. provided such funding.
19
We present intent-to-treat (ITT) effects that are unbiased estimates calculated by taking the difference
between average outcomes in the treatment and control groups.
In the appendix, we also estimate
complier average causal effects (CACEs) following the instrumental variable framework suggested by
Gerber and Green (2012) using a set of assumptions about non-compliance in the treatment group. While
we have a measure of non-compliance for respondents in the control condition (i.e., 11 percent of
respondents say that the U.S. financed the Smiling Sun Clinics without having been explicitly told so), we
do not have a measure that captures non-compliance among the treated (i.e., the proportion of
respondents who did not understand or chose not to believe the direct information about U.S. financing).
In the appendix, we make assumptions about non-compliance in the treatment condition varying from ten
17
In the control condition, 24 percent of respondents did not supply an answer, while only 17 percent of respondents in the
treatment condition declined to name a preferred donor.
18
The answer to this question, however, might be influenced by other factors besides the treatment (e.g., a preference for a foreign
country other than the United States).
19
For both of these questions, respondents almost universally assigned earnest development motives to both the United States and
Bangladesh. Slightly more than half of respondents in the treatment condition provided their thoughts on why the United States
supplied funds. With the exception of one respondent who said that the United States provided funds “for [its] own benefit,” all other
respondents who provided an answer described the United States as wanting to provide health care services or otherwise
encourage development in Bangladesh. Many more people provided a response as to why Bangladesh receives the foreign aid,
and these were nearly all development-oriented reasons. Just three respondents took the opportunity to say that it was “not good”
for Bangladesh to accept foreign money; one of these respondents related his opposition to his Islamic faith, while the other two did
not say why they felt this way.
17
to 25 percent. As is to be expected, these plausible CACE estimates are larger than the ITT estimates
that we present here.
To measure diffuse attitudes toward the United States, we told respondents that we wanted to know what
they thought “about the overall influence of some other countries on Bangladesh.” They first told us
whether they thought that a given country had a large influence or not, and among those who answered a
large influence, they told us whether it was positive or negative. The United States was one of 10
countries about which we asked. In both the treatment and control conditions, around 90 percent of
respondents said that the United States had a large influence.
The proportion was slightly (and
statistically significantly) higher in the treatment condition (δ=0.03, p < 0.01). Among those who said that
the United States had a large influence, 95 percent of respondents said that this influence was positive.
20
In the first row of Table 2, we report a difference-in-means test and a rank-sum test on a three-category
variable about U.S. influence. Given that most respondents say that the United States is influential in a
positive way, the treatment effect is small, but it is positive and obtains conventional levels of statistical
significance. In Appendix Table A4, we show that the treatment effect is driven by movement from the
“small influence” category into the “large and positive influence” category; the proportion of people
believing that the U.S. has a “large and negative influence” on Bangladesh remains constant across the
two groups.
In Table 3, we present results for this outcome variable from two sets of subgroups. First, we compare
respondents who had previous knowledge about Smiling Sun Clinics to those who did not. Second, we
compare more highly educated respondents to their less educated counterparts. The results reveal that
the treatment effect that we observe in the overall population is particularly concentrated among those
respondents who were not previously aware of the Smiling Sun Clinics and among less educated
respondents. Since respondents who were not previously aware of the Smiling Sun Clinics were also less
likely to have been aware of U.S. sponsorship of the clinics, we likely are seeing more updating of
attitudes toward the United States among this group. Among less educated respondents, it appears that
we are seeing more updating of opinions about the United States because of more negative baseline
perceptions (as indicated by the average value taken by the influence variable in the control condition).
The subgroup treatment effects, however, are not statistically significantly different from one another in
either case.
20
These high levels of pro-U.S. sentiment are comparable to recent polling by Pew Research (2014). In the Pew data, 76 percent of
Bangladeshi respondents said that they had favorable views of the United States. Among Asian countries, only respondents from
the Philippines (92 percent), South Korea (82 percent), and Vietnam (76 percent) were either equally or more positively inclined
towards the United States.
18
When respondents were asked about their preferences on specific policy relevant to the United States,
our treatment elicited less movement in attitudes. As the second row of Table 2 shows, when asked
whether “Bangladesh should put more emphasis on trading with the United States or more emphasis on
diversifying its trading partners,” respondents in the treatment condition were slightly more likely to say
that Bangladesh should trade more with the United States, but the effect was not statistically significant.
When asked which country (from a list of nine) should provide the commander for Bangladeshi troops in
U.N. Peacekeeping Operations, very few respondents in either the treatment or control condition named
the United States as the preferred country; if anything, the proportion of respondents doing so decreased
between the control and treatment conditions.
21
To measure the extent to which U.S. financing of development interventions might spill over into improved
commercial relations for U.S. companies, we gauged consumer sentiment in two ways. First, we showed
respondents logos for eight different global companies and asked respondents to rate the company on a
scale from one to seven with regard to how much they trusted the company. We included two U.S.
brands in this question: Coca-Cola and Apple. As the first two rows of Table 4 show, there were small
positive increases in the levels of trust for both companies, although only the increase for Apple, which
was greater than that for Coca-Cola and which had a lower baseline level of trust in the control condition,
was statistically significant.
We then named six different categories of consumer products and asked our respondents to name the
country that they thought did the best job of producing high quality products in that category. Overall, our
respondents did not regard the United States as a producer of high-quality consumer goods. Among
respondents who provided an answer for at least one of the categories of consumer goods, almost 70
percent did not mention the United States for any of the goods.
22
The treatment appears to have had
some marginal influence here. The third row of Table 4 demonstrates that the average number of times
the United States was mentioned as a producer of high-quality consumer goods was higher in the
treatment condition, although the difference misses conventional levels of statistical significance (p <
0.14).
In sum, we find evidence that providing information about the foreign funding of development projects to
Bangladeshi citizens leads them to update their general attitudes toward the donor and also toward a
multinational corporation associated with the donor. It is, of course, worth highlighting that Bangladeshis
21
While 29 percent of respondents said that they did not know who should be in command of a U.N. Peacekeeping Operation, the
clear favorite among those who answered the question was Malaysia, named by 57 percent of respondents who provided an
answer. India and China were the second and third most common answers.
22
Of those that did mention the United States, most commonly they cited U.S. cars. India was highly regarded as a producer of
motorcycles. China was the most highly regarded producer of televisions, refrigerators, mobile phones, and watches. Substantial
numbers of people also mentioned Malaysia as a producer of high-quality televisions and refrigerators.
19
already widely believe that the United States has a large and positive influence on their country, and so
the treatment results in some additional people saying that the United States has a large and positive
influence on Bangladesh. On specific foreign policy issues – increasing trade with the United States or
wanting Bangladeshi troops to serve under U.S. command – the treatment does not meaningfully change
Bangladeshi opinions. With regard to U.S. commercial brands, there is a slight effect of the treatment on
how people view Apple, although no effect on the already trusted Coca-Cola.
6.3 H3: Effects of Information about Foreign Aid on Domestic
Government Legitimacy
In this section, we examine how our informational treatment affects the level of confidence that citizens
express in their domestic government institutions. We also look for effects of the informational treatment
on tax morale, an attitudinal outcome that has been studied elsewhere as an indicator of the strength of
the fiscal contract between a government and its citizens (Sacks 2012; Paler 2013). Again, we present
ITT effects below and CACE estimates for different levels of possible non-compliance among the treated
in Table A7 in the Appendix. As above, the CACE results show that the effect of our treatment on the
treated is larger than the ITT effects.
Our primary outcome variables are drawn from a series of questions about the respondent’s level of
confidence in Bangladeshi organizations and institutions. Respondents were asked whether they had “full
confidence, partial confidence, or hardly any confidence” in their local government, the national
government, and their village leader, alongside other institutions such as the military, NGOs, and political
parties. In addition to studying several of these individual questions, we create a scale that ranges from 0
to 14 based on the answer to seven of the questions (i.e., central government, local government, village
leader, political parties, judiciary, military, and police). We also asked respondents to indicate the degree
to which they agreed (on a five-point scale) with the following statement: “Those who govern the country
are engaged in rampant corruption.”
Before presenting our results, it is important to highlight that the level of confidence in the national
government is quite high in general: 50 percent of the respondents indicated that they had “full
confidence,” and 13 percent said that they had “partial confidence.” Citizens’ confidence in their local
governments is somewhat lower, with 34 percent saying that they had “full confidence” and 53 percent
indicating that they had “partial confidence.” Drawing on research that explores citizen interactions with
state institutions, we know that for the daily lives of ordinary people in developing countries the central
government and its institutions remain distant and abstract entities (e.g., Bratton 2010). Citizens are more
20
likely to interact with and form political relationships with local authorities. We thus expect our treatment to
be more likely to influence citizen perceptions of their local but not necessarily their central government.
As Table 5 indicates, we find no initial evidence that our treatment undermines domestic government
legitimacy. Instead, we find evidence that our respondents express more positive opinions about their
own government in the treatment condition. Although there is no discernible difference between the
treatment and control conditions for confidence in the national government, we detect a statistically
significant increase in the levels of confidence in local government.
23
We also find that treated
respondents indicate more confidence in the overall set of state institutions captured in our additive index;
this difference is also statistically significantly. Our treatment also caused our respondents to view
politicians as less corrupt than respondents from the control group; the difference between groups is
again statistically significant.
Through a series of subgroup analyses presented in Table 6, we identify three groups for whom the
information treatment has a particularly pronounced effect on their level of confidence in the local
government: (1) previous users of the Smiling Sun Clinics, (2) people who report being active in politics,
and (3) the highly educated (i.e., those with a completed secondary education or above). For each group,
there is a plausible explanation for why we see larger treatment effects linked to these groups viewing the
development intervention as beneficial for their community. Previous users, for instance, are particularly
likely to have positive perceptions of the clinics. If being exposed to information about the foreign funding
of these clinics leads these users to perceive that their local government played a role in arranging for the
clinics to be located in their area, they will be inclined to transfer positive affect to their local government.
For politically active and more educated respondents, we similarly might expect that these individuals
perceive the difficulties of securing foreign funding in a way that makes them likely to credit the local
government for securing the Smiling Sun Clinics for their locality. As requests for health clinics originate
from local governments that lack resources for service provision, more politically active and educated
respondents may view external provision as complementary to – and not in conflict with – service
provision by the state.
When comparing previous users of the Smiling Sun Clinics to non-users, the estimated treatment effect
with regard to confidence in local government is substantially larger among previous users of the clinics.
On a three-point scale, previous users express opinions of their local government that are 0.26 points
higher in the treatment group as compared to the control group. Among non-previous users, the control
23
Appendix Table A5 shows how this result is mostly driven by movement of people who answer “partial confidence” in the control
condition into the “full confidence” category in the treatment condition.
21
and treatment groups are indistinguishable. The different treatment effects within the two subgroups are
statistically significantly different from one another (p < 0.01).
Among respondents who consider themselves at least somewhat politically active, stark differences exist
between treatment and control conditions.
24
Table 6 shows that, among politically active respondents,
our treatment increases confidence in local government by 0.22 points (p < 0.01), whereas among
respondents who do not consider themselves politically active, there is no significant difference between
the treatment and control groups in terms of confidence in the local government. Whereas our politically
active respondents have lower opinions of the local government in the control condition than nonpolitically-active respondents, the treatment improves their opinions of local government to the extent that
they are higher than those of non-politically-active respondents in the treatment condition. The difference
between the treatment effects for the two subgroups is statistically significant.
Finally, we estimate conditional average treatment effects across levels of education.
Among
respondents with secondary education and higher, the treatment improves attitudes toward the local
government by 0.16 points (p < 0.01). Among less educated respondents we do not observe a significant
improvement in attitudes. When comparing the treatment effects within these two subgroups, we find that
the difference falls below conventional levels of statistically significance (p < 0.14).
Overall, we find no evidence of a negative externality of foreign aid on domestic government legitimacy.
Instead, we find that information about the foreign sponsorship of a development intervention increases
some respondents’ confidence in their local governments. This effect is particularly concentrated in
previous users of the Smiling Sun Clinics, more politically active and higher educated respondents. It is
these subsets of people who recognize the value of the clinic and reward their local governments for
securing foreign funding for the clinics. What is more, respondents with higher levels of education and
political involvement are more likely to understand the process through which foreign aid projects are
distributed across localities and may thus be more likely to reward their local government for securing the
project.
24
Respondents could describe themselves as “not at all,” “somewhat,” or “very much” politically active.
22
7. Conclusions
Using an informational experiment embedded in a nationally representative survey in Bangladesh, we ask
three questions about how information about the foreign sponsorship of development interventions might
or might not serve its public diplomacy role of changing opinions among the citizens of an aid-receiving
country. We look first to see if the widespread technique of branding foreign aid interventions with a
donor’s logo effectively communicate that a project was funded by a foreign country. Then we look to see
whether information about external funding improves attitudes toward the donor. And finally we examine
whether there is a positive or negative externality of this information on citizens’ views about their own
government.
We find that providing information about U.S. sponsorship of a popular health intervention has a small but
detectable effect on attitudes toward the donor. From a high baseline belief that the U.S. has a large,
positive influence on Bangladesh, we find an increase in positive perceptions of the United States among
our treatment group. We also find small increases in perceptions of one U.S.-based company. We do not
find an impact on attitudes related to specific foreign policies. The estimation of heterogeneous treatment
effects reveal that our treatment is especially likely to improve attitudes toward the United States among
respondents who had previous awareness of the Smiling Sun Clinics and have at least a completed
secondary education.
We find no evidence of an oft-cited potential negative externality of non-government service provision –
that it might undermine the legitimacy of the government (and thereby undermine overall strategic
objectives of the donor). Instead, we find evidence for the opposite hypothesis: donor visibility improves
attitudes toward the local government. This effect of our information treatment is particularly pronounced
among previous clinic users, the more politically active, and respondents with higher levels of education.
Perhaps the most surprising finding in the study is how limited the information transfer from donor
branding seems to be. Immediately after watching a video branded with a donor logo, only a minority of
our respondents are able to identify the United States as the funder of the intervention portrayed in that
video. Later in the survey, we find that few respondents remember the logo. Our findings suggest that the
U.S. government’s annual multi-million dollar investments in the marking of aid with the USAID logo may
not be accomplishing their intended public diplomatic purpose. Other donor governments around the
world, including Britain and Germany, have recently changed their branding strategy to feature national
flags in lieu of logos. Whether flags are more effective brands has yet to be systematically tested. Our
23
results also suggest that donor governments may want to reconsider, in general, whether branding
through logos is a sufficiently important task in the pursuit of their national interest.
That previous users of the clinic are better able to make the association between the logo and U.S.
funding of the health clinics speaks to the ways in which repeated exposure to brands may matter. It is
among this group, as well, that we observe large effects of the treatment on attitudes toward the local
government, suggesting that information about foreign funding of development interventions will be most
salient for people who actually have had exposure to the intervention itself. That we do not see effects for
the foreign policy questions among this group also helps give us confidence in the meaningfulness of the
null effects that we observe there.
Our findings form the basis of future research. In the current survey, we have allowed respondents to
draw their own conclusions about the implications of foreign aid projects. In the real world, however,
people often rely on information sources to provide an interpretation of politically-relevant information.
That is, aid projects may be described positively or negatively by local elites and/or media. In a follow-up
survey we plan to impose information about project quality by designing survey vignettes that describe
U.S.-funded projects in positive and negative ways. If respondents are given information about projectfailure or corruption within a project, will this knowledge influence their attitudes toward the project
sponsor and their local government? Will a negative project description offset our observed positive
effects on respondent attitudes toward their local government?
Another interesting question to be addressed in future research would be to examine whether our current
project findings apply to domestic sectors beyond health. Since the project we study is specific to mothers
and newborn children and is highly valued across Bangladesh, the results may mask important
25
differences across sectors.
Respondents may evaluate government responsibilities differently in the
security sector where the state is viewed as central to the provision of order. This would require that future
evaluations would either replicate our study in the security sector or employ a new research design that
accounts for potentially important heterogeneity across foreign aid projects.
Finally, our study motivates further research that investigates the effects of donor visibility in different
political contexts. In Bangladesh, as we have shown, respondents mostly hold favorable views towards
the United States even though there is a growing current of anti-Western sentiment in the country. Even in
the face of high baseline support for the U.S., the information treatment caused respondents to regard the
United States even more positively, though the extent to which attitudes could be changed was limited. In
25
In the appendix, we look for differential effects of the treatment on women and find that treatment effects are, if anything, smaller
among this group that might have been more receptive to the information. 24
countries, like Egypt, Pakistan, or Afghanistan, pro-Americanism is notably lower, leaving more room for
people to update their beliefs. It is in these countries where donor visibility may, on balance, have a
greater effect on improving attitudes among people exposed to the treatment. Existing literature from
these contexts is inconclusive. Andrabi and Das (2010) argue that humanitarian relief helped to improve
attitudes toward foreigners in northern Pakistan, while Zürcher, Koehler, and Böhnke (2010) find null
effects of foreign assistance on attitudes toward foreigners in Afghanistan. Better understanding how
different contexts alter the effects that we have identified here will help inform both our understanding of
attitudinal change and donor policies on public diplomacy.
25
Tables and Figures
Figure 1. Screenshots from control and treatment videos
26
Table 1. Perceived origins of health clinic financing
Do you know from where the
Smiling Sun Clinics get most
of their financial assistance?
Don’t
Know
United
States
Unnamed
Foreign
Aid
Govern
-ment
Other
0.04
Specific,
Non-U.S.
Foreign
Aid
0.0009
Overall Control Group
(N=1,073)
Overall Treatment Group
(N=1,163)
0.81
0.11
0.03
0.008
0.69
0.24
0.03
0.0009
0.03
0.004
Previous Users Control Group
(N=168)
Previous Users Treatment Group
(N=198)
0.65
0.26
0.06
0.01
0.01
0.01
0.49
0.40
0.07
0.01
0.04
0.00
Previous Awareness (But Not
Use) Control Group
(N=332)
Previous Awareness (But Not
Use) Treatment Group
(N=376)
0.74
0.16
0.06
0.00
0.03
0.01
0.60
0.34
0.03
0.00
0.03
0.003
High Education Control Group
0.72
0.21
0.06
0.00
0.02
(N=251)
High Education Treatment Group 0.47
0.44
0.05
0.004
0.04
(N=263)
Notes: High education is defined as respondents with a completed secondary education or
2
value for a Pearson χ test of independence between the rows and the columns: p < 0.01
previously aware, and high education; p < 0.05 for previous users.
27
0.004
0.00
above. pfor overall,
Table 2. Responses to foreign policy questions
Outcome
Control
(Smiling
Sun Brand
Only)
Treatment
(USAID
Brand and
Information)
Difference
p-value for
H0: No
Difference
U.S. Influence: Negative, Null, or
Positive (-1, 0, 1)
0.80
(0.02)
N=1,086
0.84
(0.01)
N=1,177
0.05
(0.02)
0.03
[0.01]
Should Bangladesh trade more with
U.S. or else diversify its trading
partners? (proportion saying “trade
more”)
0.16
(0.01)
N=1,024
0.17
(0.01)
N=1,111
0.01
(0.02)
0.41
[0.40]
0.03
(0.006)
N=761
0.02
(0.005)
N=845
-0.007
(0.007)
0.42
[0.42]
Bangladeshi troops serve under the
command of military officers from other
countries in U.N. PKOs. Which
country would you want? (Proportion
of respondents with any answer saying
“United States”)
Notes: Cells present mean responses in each treatment group with standard errors in parentheses. pvalues are based on t-tests of H0: No difference in means between the treatment and control groups. pvalues in brackets are based on Wilcoxon rank sum tests.
28
Table 3. Heterogeneous treatment effects for perceived influence of U.S
U.S. Has a Good Influence on Bangladesh (-­‐1, 0, 1) Previous Awareness of SSC No Previous Awareness of SSC Completed Secondary or Above Less Than Completed Secondary Treatment (USAID Brand and Information) 0.87 (0.02) N=575 0.82 (0.02) N=602 0.88 (0.03) N=269 0.83 (0.02) N=908 Control (Smiling Sun Brand Only) 0.84 (0.02) N=500 0.03 (0.03) 0.37 [0.26] 0.75 (0.02) N=586 0.06 (0.03) 0.04 [0.02] 0.86 (0.03) N=252 0.01 (0.04) 0.77 [0.82] 0.77 (0.02) N=834 0.06 (0.02) 0.02 [0.02] Difference p-­‐value for H0: No Difference p-­‐value for H0: No Difference between CATEs 0.37 0.49 Notes: Outcome variable is coded -1 for respondents who think that the U.S. has a large and negative
influence on Bangladesh, 0 for respondents who think that the U.S. does not have an influence on
Bangladesh, and 1 for respondents who think that the U.S. has a large and positive influence on
Bangladesh. Cells present mean responses in each treatment group with standard errors in parentheses.
The p-values are based on t-tests of H0: No difference in means between the treatment and control
groups. p-values in brackets are based on Wilcoxon rank sum tests. p-value for difference between
conditional average treatment effects (CATEs) is based on a randomization-inference-based test of the
kind described in Gerber and Green (2012).
29
Table 4. Responses to commercial questions
Outcome Trust in Brand: Coca-­‐Cola (1 – 7) Trust in Brand: Apple (1 – 7) Number of six products (e.g., cars, mobile phones) for which respondents thought U.S. made “highest quality products” (0 – 6) Control (Smiling Sun Brand Only) Treatment (USAID Brand and Information) Difference p-­‐value for H0: No Difference 4.47 (0.06) N=1,086 3.60 (0.07) N=1,086 0.44 (0.03) N=972 4.53 (0.05) N=1,177 3.87 (0.06) N=1,177 0.50 (0.03) N=1,085 0.06 (0.08) 0.45 [0.49] 0.25 (0.09) 0.01 [0.02] 0.06 (0.04) 0.14 [0.22] Notes: Cells present mean responses in each treatment group with standard errors in parentheses. The
p-values are based on t-tests of H0: No difference in means between the treatment and control groups. pvalues in brackets are based on Wilcoxon rank sum tests.
30
Table 5. Responses to domestic government legitimacy questions
Outcome Level of Confidence in National Government (no, partial, or full, 1 – 3) Level of Confidence in Local Government (no, partial, or full, 1 – 3) Additive Index of Confidence in Seven Government / Political Institutions (0 – 14) Tax department has the right to make people pay taxes (strongly disagree … strongly agree, 1 – 5) Those who govern the country are engaged in rampant corruption (strongly disagree … strongly agree, 1 – 5) Control (Smiling Sun Brand Only) Treatment (USAID Brand and Information) Difference p-­‐value for H0: No Difference 2.36 (0.02) N=1,086 2.18 (0.02) N=1,086 8.37 (0.09) N=1,086 4.08 (0.03) N=1,019 3.61 (0.03) N=1,086 2.38 (0.02) N=1,177 2.24 (0.02) N=1,177 8.62 (0.09) N=1,177 4.09 (0.03) N=1,102 3.50 (0.03) N=1,177 0.01 (0.03) 0.72 [0.33] 0.07 (0.03) 0.02 [0.02] 0.25 (0.13) 0.06 [0.07] 0.00 (0.04) 0.91 [0.59] -­‐0.10 (0.05) 0.03 [0.06] Notes: Cells present mean responses in each treatment group with standard errors in parentheses. The
p-values are based on t-tests of H0: No difference in means between the treatment and control groups. pvalues in brackets are based on Wilcoxon rank sum tests.
31
Table 6. Heterogeneous treatment effects for level of confidence in local government
Level of Confidence in Local Government (no, partial, or full, 1 – 3) Completed Less Than Not Politically Secondary Completed or Above Secondary Active Previous Use of SSC No Previous Use of SSC Politically Active Treatment (USAID Brand and Information) 2.35 (0.04) N=201 2.22 (0.02) N=976 2.28 (0.04) N=192 2.23 (0.02) N=985 2.34 (0.04) N=269 2.21 (0.02) N=908 Control (Smiling Sun Brand Only) 2.09 (0.05) N=170 0.26 (0.06) 0.01 [0.01] 2.19 (0.02) N=916 0.03 (0.03) 0.40 [0.32] 2.06 (0.05) N=185 0.22 (0.06) 0.01 [0.01] 2.20 (0.02) N=901 0.03 (0.03) 0.26 [0.19] 2.18 (0.04) N=252 0.16 (0.06) 0.01 [0.01] 2.18 (0.02) N=834 0.04 (0.03) 0.25 [0.20] Difference p-­‐value for H0: No Difference p-­‐value for H0: No Difference between CATEs 0.01 0.01 0.14 Notes: Cells present mean responses in each treatment group with standard errors in parentheses. The
p-values are based on t-tests of H0: No difference in means between the treatment and control groups. pvalues in brackets are based on Wilcoxon rank sum tests. p-value for difference between conditional
average treatment effects (CATEs) is based on a randomization-inference-based test of the kind
described in Gerber and Green (2012).
32
Appendix
Sampling Strategy
The primary sampling unit (PSU) for the survey was the mouza, an administrative unit roughly equivalent
to a village. We randomly sampled 96 mouza using a probability-proportional-to-size weighting based on
the 2011 population census. The 96 mouza cover 53 out of 64 districts in Bangladesh. For each PSU, 25
households were selected using a random walk where every tenth household was chosen. Within each
household, a Kish grid was used to identify a respondent over the age of 18. Not all sampled units were
eventually surveyed, resulting in a total N of 2,294, rather than the anticipated N of 2,400.
Table A1. Balance Checks
Variable
Female (0/1)
Age
Education (7-point scale)
Minority Group (0/1)
Importance of Health Care as a
Policy Issue (4-point scale)
Previously Used Smiling Sun
Clinic (0/1)
Previously Aware of Smiling Sun
Clinics (0/1)
Control Group
Mean (SE)
Treatment
Group Mean
(SE)
0.51
(0.01)
N=1,167
37.9
(0.41)
N=1,177
3.92
(0.06)
N=1,177
0.15
(0.01)
N=1,177
3.81
(0.01)
N=1,174
0.17
(0.01)
N=1,177
0.49
(0.01)
N=1,177
0.54
(0.02)
N=1,076
37.6
(0.42)
N=1,086
3.82
(0.06)
N=1,086
0.17
(0.01)
N=1,086
3.80
(0.01)
N=1,082
0.16
(0.01)
N=1,086
0.46
(0.02)
N=1,086
33
Difference (SE)
p-value for H0:
No Difference
0.03
(0.02)
0.13
-0.29
(0.59)
0.62
-0.10
(0.09)
0.25
0.02
(0.02)
0.27
-0.02
(0.02)
0.30
-0.01
(0.02)
0.37
-0.03
(0.02)
0.19
Knowledge of U.S. Funding in the Control Group
Among the control group, we run a regression predicting whether or not the respondents correctly
answered the question about the funding of the Smiling Sun Clinics by identifying the United States as the
funder.
Table A2. Control Group Knowledge of U.S. Funding
U.S. Provided Funding for
SSCs
Previous Clinic Users
(1)
Previously Aware of the Clinic
Female
Age
Education Level
Minority Group Status
Income
Subjective Quality of Life
Assessment
Access to a Television
1.02***
(0.30)
0.88***
(0.28)
-0.53**
(0.24)
-0.02*
(0.01)
0.16**
(0.07)
0.66**
(0.30)
0.06
(0.05)
0.42*
(0.23)
0.29
(0.29)
Upazilla Fixed Effects
Y
N
970
2
Pseudo-R
0.15
Notes: Results from a logistic regression. *** - p < 0.01; ** - p < 0.05; * - p < 0.10.
34
Regression-Based Tests
In the paper, we rely on difference-in-means tests for all of our results. Here we look for treatment effects
in regression models that control for socio-demographic characteristics of our respondents. Specifically,
we control for whether or not the respondent is a woman, age, education, whether or not the respondent
belongs to an ethnic minority group, self-reported household income, the respondent’s subjective
assessment of their family’s living conditions, an indicator for whether or not the respondent has access to
a television, and upazilla (subdistrict) fixed effects.
The regression models return coefficients of similar magnitude and levels of statistical significance to the
difference-in-means tests reported in the main text.
35
Outcome
Recognition of U.S. Brand
Identify U.S. as Funder of Smiling Sun Clinics (0/1)
Foreign Policy Outcomes
U.S. Influence: Negative, Null, or Positive (-1, 0, 1)
Should Bangladesh trade more with U.S. or else
diversify its trading partners? (0/1)
Bangladeshi troops serve under the command of
military officers from other countries in U.N. PKOs.
Which country would you want? (For respondents
with any answer, whether they said “United
States”, 0/1)
Commercial Outcomes
Trust in Brand: Coca-Cola (1-7)
Type of
Regression
Model
Coefficient on
Treatment
Indicator
(S.E.)
N
Logistic
0.97***
(0.13)
2,243
Ordered
Logistic
Logistic
0.33***
(0.13)
0.12
(0.12)
-0.40
(0.35)
2,243
-0.004
(0.08)
0.15**
(0.08)
0.15
(0.10)
2,243
Logistic
Ordered
Logistic
Ordered
Logistic
Ordered
Logistic
Trust in Brand: Apple (1-7)
2,117
1,594
2,243
Number of Six Products for Which Respondent
2,039
Thought that U.S. Made “High Quality” Goods (0 6)
Domestic Government Legitimacy
Level of Confidence in National Government (no,
Ordered
0.08
2,243
partial, or full, 1 – 3)
Logistic
(0.08)
Level of Confidence in Local Government (no,
Ordered
0.20**
2,243
partial, or full, 1 – 3)
Logistic
(0.08)
Additive Index of Confidence in Seven
Ordered
0.14*
2,243
Government / Political Institutions (0 – 14)
Logistic
(0.07)
Tax department has the right to make people pay
Ordered
0.04
2,103
taxes (strongly disagree … strongly agree, 1 – 5)
Logistic
(0.09)
Those who govern the country are engaged in
Ordered
-0.14*
2,243
rampant corruption (strongly disagree … strongly
Logistic
(0.07)
agree, 1 – 5)
Table A3. Regression-Based Analyses. Coefficient and standard error from (ordered) logistic
regression models including an indicator for female respondents, age, education, an indicator for ethnic
minority status, household income, subjective assessment of living conditions, an indicator for access to a
television, and upazila (sub-district) fixed effects.
36
Specific Movement in Outcome Variables
Table A4. Cross-Tabs of U.S. Influence Variable and Treatment
Control
Treatment
Large and Negative Influence
0.05
0.04
Small Influence
0.11
0.07
Large and Positive Influence
0.84
0.88
2
Notes: Proportion of cases in each cell. p < 0.02 in a χ test of independence between rows and columns.
Table A5. Cross-Tabs of Local Government Confidence Variable and Treatment
Control
Treatment
No Confidence
0.14
0.13
Partial Confidence
0.55
0.50
Full Confidence
0.31
0.37
2
Notes: Proportion of cases in each cell. p < 0.02 in a χ test of independence between rows and columns.
37
Estimation of Complier Average Causal Effects
We estimate complier average causal effects (CACEs) based on observed and assumed patterns of noncompliance. In Table 2 in the paper we provide an estimate of non-compliance in the control group: 11
percent of the respondents in the control group say that the United States funds the SSCs, even though
this group had not been given the information that identified the United States as sponsor of the
intervention. Although our survey instrument included a manipulation check for the identification of the
logo, we did not include the same check for the information treatment and are thus unable to present
exact numbers as to how many respondents in the treatment group were unable to identify the United
States as the SSC sponsor. Therefore, we estimate a series of CACEs assuming different proportions of
non-compliance (e.g., 10%, 15%, 20%, 25%, 30%, and 35%) in the treatment group.
Using the CACE theorem for two-sided non-compliance (see Gerber and Green 2014, p. 183-186), we
estimate the CACE by dividing the ITT by the difference in the proportions of compliers in treatment and
control groups. The CACE provides consistent estimates of the average treatment effect among
compliers.
26
In Tables A6 and A7 we show CACEs across six assumed proportions of non-compliers for
our measure that captures positive attitudes toward the United States and our measure of trust in local
government.
Table A6 lists the estimates across assumed levels of non-compliance for respondent positive attitudes
toward the United States. Given the non-compliers in the control group, the size of the CACE increases
slightly as we assume more and more non-compliers in the treatment group.
27
Across different
proportions of non-compliance among the treated, the average treatment effect among compliers is
statistically significant. For example, if we assume that 10 percent of the respondents in the treatment
group did not actually receive the treatment, we would expect our information treatment to improve
perceptions of the United States among those who did receive the treatment by 0.06 points (on a threepoint scale). If we assume that 20 percent of respondents did not receive the treatment, then we estimate
the CACE to be a 0.07 point improvement. These estimates suggest that even among those who fully
absorbed the information about U.S. funding, the extent to which we can observe more positive attitudes
about the United States is constrained – a result of the high levels of baseline belief that the United States
has a positive influence on Bangladesh.
26
We estimate the CACE under assumptions of monotonicity, non-interference, and excludability.
This relationship holds as long as the proportion of treatment compliers is greater than the proportion of non-compliers in the
control. 27
38
Table A6. CACEs for Perceived Influence of U.S.
Outcome: Perceived Influence of U.S.
CACE
Robust
Assuming the following proportion of
Std. Error
non-compliance within the treatment
group…
0.10
0.060
0.262
0.15
0.064
0.280
0.20
0.068
0.030
0.25
0.074
0.032
0.30
0.081
0.035
0.35
0.089
0.038
Notes: Outcome variable is coded -1 for respondents who think that the U.S. has a large and negative
influence on Bangladesh, 0 for respondents who think that the U.S. does not have an influence on
Bangladesh, and 1 for respondents who think that the U.S. has a large and positive influence on
Bangladesh. Coefficient and standard error are based on 2SLS estimates described in Gerber and Green
(2012).
Table A7 lists the estimates across assumed levels of non-compliance for respondent’s trust in their local
government. Given the non-compliers in the control group, the size of the CACE (again) increases as we
increase the assumed proportion of non-compliers in the treatment group. For example, if we assume that
10 percent of the respondents in the treatment group did not actually receive the treatment, we would
expect our information treatment to increase trust in the local government by 0.08 points. If we assume
that 20 percent of respondents did not receive the treatment, then we estimate the CACE to be a 0.09
point improvement.
Table A7. CACEs for Perceived Trust in Local Government
Outcome: Perceived Trust in Local Gov’t
CACE
Robust
Assuming the following proportion of
Std. Error
non-compliance within the treatment
group…
0.10
0.082
0.349
0.15
0.088
0.373
0.20
0.095
0.040
0.25
0.102
0.043
0.30
0.111
0.046
0.35
0.121
0.051
Notes: Coefficient and standard error are based on 2SLS estimates described in Gerber and Green
(2012).
39
Treatment Effects among Women
Given the substantive content of the video, we might expect that women might be more sensitive to the
information about the funding of the Smiling Sun Clinics. The results reveal that, if anything, our treatment
effects are concentrated among our male respondents. Treatment effects among women are often
smaller and less likely to be statistically significant.
Table A8. Replication of Table 1 (Perceived Origins of Health Clinic Financing) for Women Only
Do you know from where the
Smiling Sun Clinics get most
of their financial assistance?
Don’t
Know
United
States
Unname
d
Foreign
Aid
Government
Other
0.04
Specific,
NonU.S.
Foreign
Aid
0.002
Overall Control Group
(N=578)
Overall Treatment Group
(N=587)
0.81
0.11
0.03
0.009
0.71
0.21
0.03
0.002
0.04
0.005
Previous Users Control Group
(N=93)
Previous Users Treatment Group
(N=117)
0.63
0.29
0.05
0.01
0.01
0.00
0.50
0.35
0.09
0.01
0.05
0.00
Previous Awareness (But Not
Use) Control Group
(N=164)
Previous Awareness (But Not
Use) Treatment Group
(N=172)
0.78
0.12
0.05
0.05
0.04
0.01
0.67
0.26
0.02
0.02
0.05
0.01
High Education Control Group
0.70
0.22
0.04
0.00
0.04
0.00
(N=114)
High Education Treatment Group
0.48
0.37
0.07
0.01
0.06
0.00
(N=97)
Notes: High education is defined as respondents with a completed secondary education or above. p2
value for a Pearson χ test of independence between the rows and the columns: p < 0.01 for overall; p <
0.25 for previous users; p < 0.03 for previously aware; and p < 0.03 for high education.
40
Table A9. Replication of Table 2 (Responses to Foreign Policy Questions) for Women Only
Outcome
U.S. Influence: Negative, Null, or
Positive (-1, 0, 1)
Control
(Smiling
Sun Brand
Only)
Treatment
(USAID
Brand and
Information
)
Difference
p-value for
H0: No
Difference
0.79
(0.02)
N=583
0.83
(0.02)
N=595
0.04
(0.03)
0.14
[0.09]
Should Bangladesh trade more with
0.18
0.18
0.00
0.99
U.S. or else diversify its trading
(0.02)
(0.02)
(0.02)
[0.99]
partners? (proportion saying “trade
N=542
N=547
more”)
Bangladeshi troops serve under the
0.03
0.00
-0.03
0.01
command of military officers from other
(0.009)
(0.00)
(0.008)
[0.01]
countries in U.N. PKOs. Which
N=359
N=394
country would you want? (Proportion
of respondents with any answer saying
“United States”)
Notes: Cells present mean responses in each treatment group with standard errors in parentheses. pvalues are based on t-tests of H0: No difference in means between the treatment and control groups. pvalues in brackets are based on Wilcoxon rank sum tests.
Table A10. Replication of Table 4 (Responses to Commercial Questions) for Women Only
Outcome
Trust in Brand: Coca-Cola (1 – 7)
Trust in Brand: Apple (1 – 7)
Control
(Smiling
Sun Brand
Only)
Treatment
(USAID
Brand and
Information
)
Difference
p-value for
H0: No
Difference
4.37
(0.08)
N=583
3.49
(0.09)
N=583
0.43
(0.04)
N=501
4.49
(0.08)
N=595
3.68
(0.09)
N=595
0.50
(0.03)
N=529
0.12
(0.11)
0.25
[0.25]
0.19
(0.12)
0.13
[0.16]
Number of six products (e.g., cars,
0.07
0.27
mobile phones) for which respondents
(0.06)
[0.31]
thought U.S. made “highest quality
products” (0 – 6)
Notes: Cells present mean responses in each treatment group with standard errors in parentheses. The
p-values are based on t-tests of H0: No difference in means between the treatment and control groups. pvalues in brackets are based on Wilcoxon rank sum tests.
41
Table A11. Replication of Table 5 (Responses to Domestic Government Legitimacy Questions)
for Women Only
Outcome
Level of Confidence in National
Government (no, partial, or full, 1 – 3)
Level of Confidence in Local
Government (no, partial, or full, 1 – 3)
Control
(Smiling
Sun Brand
Only)
Treatment
(USAID
Brand and
Information
)
Difference
p-value for
H0: No
Difference
2.40
(0.03)
N=583
2.19
(0.03)
N=583
8.50
(0.12)
N=583
4.09
(0.04)
N=537
3.63
(0.04)
N=583
2.39
(0.03)
N=595
2.23
(0.03)
N=595
8.64
(0.12)
N=595
4.06
(0.04)
N=542
3.56
(0.05)
N=595
-0.01
(0.04)
0.81
[0.87]
0.04
(0.04)
0.31
[0.30]
Additive Index of Confidence in Seven
0.14
0.43
Government / Political Institutions (0 –
(0.17)
[0.52]
14)
Tax department has the right to make
-0.03
0.59
people pay taxes (strongly disagree …
(0.05)
[0.99]
strongly agree, 1 – 5)
Those who govern the country are
-0.07
0.27
engaged in rampant corruption
(0.06)
[0.34]
(strongly disagree … strongly agree, 1
– 5)
Notes: Cells present mean responses in each treatment group with standard errors in parentheses. The
p-values are based on t-tests of H0: No difference in means between the treatment and control groups. pvalues in brackets are based on Wilcoxon rank sum tests.
42
References
Adelman, Ken. 2011. “Not-So-Smart Power.” Foreign Policy.
http://www.foreignpolicy.com/articles/2011/04/18/not_so_smart_power (accessed 13 August
2013).
Anam, Tahmima. 2015. “Bangladesh on the Brink,” New York Times 4 November,
http://www.nytimes.com/2015/11/05/opinion/bangladesh-on-the-brink.html (accessed 5 November
2015).
Andrabi, Tahir, and Jishnu Das. 2010. In Aid We Trust: Hearts and Minds and the Pakistan Earthquake of
2005. Washington, D.C.: The World Bank.
Bakke, Kristin. 2015. External Patrons, Violence, and Internal Legitimacy in de facto States: Abkhazia,
Nagorno Karabakh, South Ossetia, and Transdniestria Compared.” Unpublished manuscript.
Bearce, David, Steven E. Finkel, Anibal S. Pérez-Liñán, Juan Rodríguez-Zepeda, and Lena SurzhkoHarned. 2013. “Has Aid for Trade Increased Recipient Exports? The Impact of U.S. Aid
Allocations 1999-2008.” International Studies Quarterly 57(1): 163-70.
Bearce, David, and Daniel Tirone. 2010. Foreign Aid Effectiveness and the Strategic Goals of Donor
Governments. Journal of Politics 72(3): 837-851.
Berman, Elie, Jacob N. Shapiro, and Joseph H. Felter. 2011. “Can Hearts and Minds Be Bought? The
Economics of Counterinsurgency in Iraq.” Journal of Political Economy 119 (4): 766-819.
Birdsall, Nancy, Milan Vaishnav, and Daniel Cutherell. 2012. More Money, More Problems: A 2012
Assessment of the US Approach to Development in Pakistan. Washington, DC: Center for Global
Development.
http://www.cgdev.org/files/1426361_file_CGD_More_Money_More_Problems_web.pdf (August
13, 2013).
Blair, Robert. 2015. “UN Peacekeeping and State Legitimacy.” Unpublished Manuscript.
Böhnke, Jan Rasmus, and Christoph Zürcher. 2013. “Aid, Minds and Hearts: The Impact of Aid in Conflict
Zones.” Conflict Management and Peace Science 30(5): 411–32.
Brass, Jennifer N. 2010. “Surrogates for Government? NGOs and the State in Kenya.” University of
California, Berkeley.
http://c.ymcdn.com/sites/www.istr.org/resource/resmgr/siena_dissertations/brass_dissertation.pdf
?hhSearchTerms=%22national+and+campaigns+and+charity%22 (February 2, 2015).
Bratton, Michael. 1989. “The Politics of Government-NGO Relations in Africa.” World Development 17(4):
569–87.
43
Bratton, Michael. 2010. Citizen Perceptions of Local Government Responsiveness in Sub-Saharan Africa.
Afrobarometer Working Paper 119.
http://www.eldis.org/vfile/upload/1/document/1109/AfropaperNo119_2.pdf
Brautigam, Deborah. 2000. Aid Dependence and Governance. Washington, D.C.: American University.
http://www1.american.edu/faculty/brautigam/Aid%20Dependence%20and%20Governance.pdf
(September 24, 2009).
Cattin, Philippe J., Alain Jolibert, and Colleen Lohnes. 1982 “A Cross-Cultural Study of ‘Made in’
Concepts.” Journal of International Business Studies 13 (3): 131–141.
Carnegie, Allison, and Lindsay Dolan. 2015. “The Effects of Aid on Recipients’ Reputations: Evidence
from Natural Disaster Response. Paper Presented at the ISA Conference “Foreign Aid and
Government Legitimacy.” Washington, D.C.
Cruz, Cesi, and Christina J. Schneider. Forthcoming. “Foreign Aid and Undeserved Credit Claiming,”
American Journal of Political Science.
Department for International Development (U.K.). 2012. New Logo: Flying the Flag for UK Aid. London.
https://www.gov.uk/government/news/new-logo-flying-the-flag-for-uk-aid (March 18, 2015).
Dietrich, Simone, and Matthew S. Winters. 2015. “Foreign Aid and Government Legitimacy.” Journal of
Experimental Political Science.
Dietrich, Simone, and Joseph Wright. 2015. “Foreign Aid Allocation Tactics and Democratic Change in
Africa.”
Fishstein, Paul, and Andrew Wilder. 2012. Winning Hearts and Minds? Examining the Relationship
between Aid and Security in Afghanistan. Medford, MA: Feinstein International Center, Tufts
University. http://www.humanitarianresponse.info/fr/system/files/documents/files/WinningHeartsFinal.pdf (March 14, 2015).
Flores, Thomas, and Irfan Nooruddin. 2009. “Financing the Peace: Evaluating World Bank Post-conflict
Assistance Programs. Review of International Organizations 4(1): 1-27.
Fowler, Alan. 1991. “The Role of NGOs in Changing State-Society Relations: Perspectives from Eastern
and Southern Africa.” Development Policy Review 9(1): 53–84.
Gerber, Alan S., and Donald P. Green. 2012. Field Experiments: Design, Analysis, and Interpretation.
New York: W.W. Norton & Co.
Goldsmith, Benjamin E., Yusaku Horiuchi, and Terence Wood. 2014. “Doing Well by Doing Good: The
Impact of Foreign Aid on Foreign Public Opinion.” Quarterly Journal of Political Science 9: 87–
114.
Gubser, Peter. 2002. “The Impact of NGOs on State and Non-State Relations in the Middle East.” Middle
East Policy 9(1): 139–48.
44
Guiteras, Raymond, and Ahmed Mushfiq Mobarak. 2014. “Does Development Aid Undermine Political
Accountability? Leader and Constituent Responses to a Large-Scale Intervention.” University of
Maryland. http://sites.bu.edu/neudc/files/2014/10/paper_488.pdf (March 12, 2015).
Han, C. Min, and Vern Terpstra. “Country-Of-Origin Effects for Uni-National and Bi-National Products.”
Journal of International Business Studies 19, no. 2 (1988): 235–255.
Hoffman, Barak D., and Clark C. Gibson. 2005. “Fiscal Governance and Public Services: Evidence from
Tanzania and Zambia.” San Diego: Department of Political Science, University of California, San
Diego. http://people.wm.edu/~mjtier/intlpolitics/seminar/papers/Fiscal%20Governance.pdf
(August 12, 2013).
Hyde, Susan. 2015. “Experiments in International Relations: Lab, Survey, and Field.” Annual Review of
Political Science. 18: 403-424.
Katzenstein, Peter J., and Robert O. Keohane, eds. 2007. Anti-Americanism in World Politics. Ithaca, NY:
Cornell University Press.
Knack, Stephen. 2001. “Aid Dependence and the Quality of Governance: Cross-Country Empirical Tests.”
Southern Economic Journal 68(2): 310–29.
Lederer, William J., and Eugene Burdick. 1958. The Ugly American. New York: W.W. Norton & Co.
Lodge, Milton, and Charles S. Taber. 2013. The Rationalizing Voter. New York: Cambridge University
Press.
Lyall, Jason, and Isaiah Wilson. 2009. “Rage Against the Machines: Explaining Outcomes in
Counterinsurgency Wars. International Organization 63: 67-106.
Merkovic-Orenstein, Alex. 2011. “Stop Trying to Win Hearts and Minds.” Partisans.
http://www.partisans.org/node/808 (August 13, 2013).
Milner, Helen V., Daniel L. Nielson, and Michael Findley. 2013. Which Devil in Development? A
Randomized Study of Citizen Actions Supporting Foreign Aid in Uganda. Rochester, NY: Social
Science Research Network. SSRN Scholarly Paper. http://papers.ssrn.com/abstract=2134409
(August 12, 20oo13).
Narang, Neil. 2013. “Biting the Hand that Feeds: An Organizational Theory Explaining Attacks Against Aid
Workers in Civil War, 1997-2010. Paper presented at the Niehaus Center for Globalization and
Governance Conference on “Research Frontiers in Foreign Aid.” Princeton University.
Nooruddin, Irfan. 2014. “Making Surveys Work Better: Experiments in Public Opinion Research.” Studies
in Indian Politics 2(1): 105-108
Nye, Joseph S. 2002. The Paradox of American Power: Why the World’s Only Superpower Can’t Go It
Alone. New York: Oxford University Press.
Paler, Laura. 2013. “Keeping the Public Purse: An Experiment in Windfalls, Taxes, and the Incentives to
Restrain Government.” American Political Science Review 107(04): 706–25.
Peabody, Dean. National Characteristics. Cambridge: Cambridge University Press, 1985.
45
Sacks, Audrey. 2012. Can Donors and Non-State Actors Undermine Citizens’ Legitimating Beliefs?.
Washington, D.C.: World Bank. Policy Research Working Paper.
Shah, Rajiv. 2010. "From the American People." Huffington Post. October 22, 2010.
http://www.huffingtonpost.com/dr-rajiv-shah/from-the-american-people_1_b_772736.html
Simon Reid-Henry. 2011. “Why Western Aid Workers Are Coming under Threat | Simon Reid-Henry |
Global Development | Guardian.co.uk.” Poverty Matters Blog. http://www.guardian.co.uk/globaldevelopment/poverty-matters/2011/may/27/western-aid-workers-under-threat (September 9,
2012).
Slovic, P. and E. Peters. 2006. “Risk Perception and Affect.” Current Directions in Psychological Science
15(6): 322-325.
Steinwand, Martin. 2015. “Compete or Coordinate? Aid Fragmentation and Lead Donorship.” International
Organization 69(2): 443-472.
Steinwand, Martin. 2014. Foreign Aid and Political Stability. Conflict Management and Peace Science.
doi:10.1177/0738894214541227. In print.
Swedlund, Haley. 2014. “Dicey Donors: How Donor Commitment Problems Can Explain Changing
Foreign Aid Practices.” Paper Presented at the American Political Science Association Annual
Meeting, Washington D.C.
U.S. Agency for International Development (USAID). 2005. Graphic Standards Manual. Washington, D.C.:
USAID.
http://www.usaid.gov/sites/default/files/documents/1869/USAID_Graphic_Standards_Manual.pdf
(February 2, 2015).
Whaites, Alan. 1998. “NGOs, Civil Society and the State: Avoiding Theoretical Extremes in Real World
Issues.” Development in Practice 8(3): 343–49.
Wilson, Robyn, and Joseph Arvai. 2006. “When Less is More: How Affect Influences Preferences When
Comparing Low and High-Risk Options.” Journal of Risk Research 9(2): 165-178.
Wilder, Andrew. 2010. “Aid and Stability in Pakistan: Lessons from the 2005 Earthquake Response.”
Disasters 34: 406–426. doi:10.1111/j.1467-7717.2010.01209.x
Zürcher, Christoph, Jan Koehler, and Jan Böhnke. 2010. Assessing the Impact of Development
Cooperation in North East Afghanistan 2005-2009. Berlin: German Federal Ministry for Economic
Cooperation and Development. http://www.oecd.org/countries/afghanistan/46785983.pdf (March
14, 2015).
46